CAN TECHNOLOGY RESHAPE AMERICA’S ELECTION SYSTEM?

Knowledge Sharing Article © 2020 Dell Inc. or its subsidiaries.

CAN TECHNOLOGY RESHAPE AMERICA’S ELECTION SYSTEM?

Bruce [email protected]

2021 Dell Technologies Proven Professional Knowledge Sharing 2

The Dell Technologies Proven Professional Certification program validates a wide range of skills

and competencies across multiple technologies and products.

From Associate, entry-level courses to Expert-level, experience-based exams, all professionals

in or looking to begin a career in IT benefit from industry-leading training and certification paths

from one of the world’s most trusted technology partners.

Proven Professional certifications include:

• Cloud

• Converged/Hyperconverged Infrastructure

• Data Protection

• Data Science

• Networking

• Security

• Servers

• Storage

• Enterprise Architect

Courses are offered to meet different learning styles and schedules, including self-paced On

Demand, remote-based Virtual Instructor-Led and in-person Classrooms.

Whether you are an experienced IT professional or just getting started, Dell Technologies

Proven Professional certifications are designed to clearly signal proficiency to colleagues and

employers.

Learn more at www.dell.com/certification

dell.com/certification 3

Table of Contents

General Election Architecture ..................................................................................................... 5

Voter Registration Database ................................................................................................... 7

Purging Data Records ........................................................................................................10

Poll Books .............................................................................................................................12

Casting a Vote .......................................................................................................................13

Verifying the Signature .......................................................................................................17

Vote Tabulation .....................................................................................................................19

The Importance of Auditing .......................................................................................................21

Voter Fraud ...............................................................................................................................22

Future Voting Technology .........................................................................................................28

New Voting Equipment ..........................................................................................................32

The Intersection of Cryptography and Mathematics May Hold the Answer .........................32

Conclusion ................................................................................................................................34

List of Abbreviations ..................................................................................................................36

Footnotes ..................................................................................................................................37

Disclaimer: The views, processes or methodologies published in this article are those of the

author. They do not necessarily reflect Dell Technologies’ views, processes or methodologies.


Democracies are built upon the guarantee of free and fair elections, and in the United States,

most candidates are elected to office by popular vote. Voting system technology seems simple,

but it is anything but that. Those in favor of computer-assisted voting include election officials

who appreciate the efficiency offered by electronic voting and citizens who enjoy smartphone

simplicity. Skeptics, including IT experts and other citizen groups, are concerned about the

safeguards that locally assembled and operated highly decentralized systems offer candidates.

Many of the election system challenges are attributable to the U.S. principle of states’ rights.

There is little agreement on the methods or ballots used by over 160 million voters who live in

over 10,000 election jurisdictions of states, territories, villages, towns, cities, counties, districts,

and areas.1,2,3 As long as states meet their Constitutional obligation to hold the election, they are

generally free to run them as they see fit. They design their databases, create rules for mail-in

ballots, decide on in-person ID requirements, and other implementation issues.4 For example,

Wisconsin alone has 1,850 cities, towns, and villages that run their own elections.5

Leading up to the 2020 election, voters felt e-voting

could improve access and ease administration.

However, the pandemic put a wrench in the e-voting

effort and it became a topic of debate. As a result,

paper ballots made a comeback, harkening back to the

days of Grover Cleveland’s 1892 Presidential election.6

Every four years, according to the Constitution’s Article II, Section 1, the presidential choice is

decided by a group of Electoral College “electors” and not directly by citizens. Each state gets a

minimum of three electors based on two Senators and at least one Congressperson. The

College has a current maximum of 538 electors. In 48 states and Washington, D.C., the top

vote-getter gets all the state allocated electoral votes, while Maine and Nebraska allocate as a

percentage of the popular vote. A candidate earning 270 electoral votes is announced in mid-

December as the next President and sworn into office on January 20 of the following year.

While the popular vote on election night often signals who will be the new President, in 2016,

Hillary Clinton received nearly 3 million more votes than Donald Trump (2.1% difference) but

lost the electoral college vote 227 to 304 giving the presidency to Trump.

As systems become complex, they can become prone to unintentional error and subject to

manipulation. In the IT world, safeguards such as firewalls and backups protect against

something going wrong or having severe implications. The most common form of election


safeguard is an auditable paper trail. With so many votes cast, even the slightest percentage of

error when counting ballots can impact an outcome.

In the time between elections, the Voter Registration Database (VRD) must be maintained,

equipment purchased, staff trained and coordinated, and voter education materials created. In

the days, weeks, and months leading up to an election, these paper and electronic systems

have to be designed, configured, and/or programmed to allow the voter to easily state their

intention and for administrators to accurately record those intentions on Election Day.

This paper is about the challenges faced by IT voting systems and what can be done to improve

them. NOTE: The election community’s rich set of acronyms are listed at the end of the paper.

General Election Architecture

With a myriad of election systems and

configurations found in the United States,

this section focuses on the basic “Voter

Registration”, “Vote Capture”, and “Vote

Tabulation” components.7

Many jurisdictions leverage computer technology to run an election and begin the process with

a supplier’s overarching Voter Management System (VMS). A VMS contains guided methods

to help election officials with candidate nominations. Based on a VRD of eligible voters, it also

assists with ballot creation, printing, addressing, and mailing. The VMS helps program precinct

equipment, coordinate logistics for polling locations

and workers, tabulate precinct and central

mail-in results, and final results reporting.

For the most part, components utilize Commercial-Off-The-Shelf (COTS) packaged products

such as standard Dell servers, hyperconverged platforms, Oracle or Microsoft databases, Cisco

networking equipment, VMware and Hyper-V virtualization, AWS cloud implementations, and so

forth, allowing engineers to focus on election design elements while relying on commercial

hardware and software support. Depending on a jurisdiction's level of modernization,

component duplication at the state and local level can provide backup or disaster recovery.

While state and county election boards try to have a modern infrastructure, older gear is often

commonplace as replacement funding is often lacking. The following general Voting System

Workflow depicts some of the processes of running an election in one jurisdiction.8

Voter Management System

ElectionConfiguration

ResultsManagement

Election Night Reporting

Warehouse Logistics

ElectionPlanning


Funding greatly impacts the resources needed to keep up with voter demand on days leading

up to and on Election Day, often manifesting itself in long

precinct lines. In Texas, Harris County serves Houston’s ~2.4

million voters with over 8,000 HART eSlate Direct-Recording

Electronic (DRE) voting machines while Dallas County’s 1.3

million voters use 4,000 ES&S ExpressVote Ballot Marking

Devices.9 These counties exemplify the various systems in simultaneous use on Election Day.

There are numerous subsystems in a typical election system, and jurisdictions tend to

implement these interrelated functions differently.10 Election Day is typically the culmination of

twelve months of preplanning. This illustration

shows the three major phases of election

preparedness, with much of the functionality

focused on the next election, while alternative

voting handles absentee ballots typically after

the polling place operations close.

Jurisdictions that permit early voting, perhaps

weeks before the official Election Day, are part of the Election Day process.

2018 Maker Type Model QTY

HART DRE ESLATE 8,189

HART DRE JBC 2,072

HART DRE DAU 1,940

HART Scanner Kodak 1-660 8

ES&S Ballot Mark ExpressVote 4,000

ES&S Scanner DS200 1,000

ES&S DRE DS850 2

Harris

County

Dallas

County


Voter Registration Database

The Constitution of 1789 does not specify who can vote.11 That job was left to individual states,

and hundreds of years ago, that generally meant white male landowners 21 and older could

vote. These days, every U.S. citizen 18 years and older that meets individual state regulations is

eligible to vote after registering through a state or local registration system.

The VRD is just one piece of the registration system. Elections are complex and often present

themselves as a logistical nightmare. There are roughly 330 million Americans and 75% of them

are 18 years and older. Of those eligible, ~64% register to vote and appear in a local VRD.12 A

VRD has many uses, among them is to create an accurate list of those ~160 million voters. It

also contains other federally suggested election metadata as implemented by state and local

entities such as a voter signature and other identification that permits voters to vote at a polling

place or by mail.

The U.S. has many individual VRDs in use. In 2016, 38 states had individual systems. In Texas,

215 counties use the state system and 39 counties had their own VRD.13 In the U.S., these

systems support more than 10,000 voting jurisdictions and 1.4 million poll workers with over

800,000 voting machines that in some way capture a vote as shown below.14,15,16 Not only can

equipment vary by jurisdiction, the basic functionality of the

equipment can differ. For example, some states using DREs

can produce a paper audit trail while others cannot, and other

jurisdictions support DRE machines and paper ballots.

The registration database’s primary use is to generate paper and electronic poll books of voters

allowed at a particular precinct and for absentee/mail-in ballot processing. It includes a voter’s

address, registration form signature, and whether they already submitted a ballot. Some

jurisdictions update or journal the database’s signature entry when a voter signs the poll book.

The VRD also maintains political party affiliation to aid officials in staffing primary elections.

The general steps of processing a mail-in or absentee ballot include:17

1. Ballots received are pre-processed, preparing them for counting before Election Day. Wisconsin, Pennsylvania, and parts of Michigan begin on Election Day.

2. Barcoded outer envelopes are scanned and checked to see if the voter already voted. 3. Jurisdictions with outer envelope signatures and addresses are cross-checked. 4. “Problem envelopes” can be submitted to a cure process allowing for voter remediation. 5. Outer envelopes are sorted by precinct and can be alphabetized for a VRD check. 6. The ballot is removed from the outer (and optional inner) envelope by hand or machine. 7. Ballots are flattened and examined before scanning, trying to prevent scanner jams. 8. Salvageable damaged ballots can be hand-transcribed to a fresh ballot.

States and Washington D.C. # States

DRE with/without paper trail 1

DRE without paper trail 4

Mail 3

DRE/Paper with/without paper trail 2

DRE/Paper with paper trail 16

DRE/Paper DRE without paper trail 7

Paper Ballot 18

51


9. Ballots are scanned. 10. The results are tabulated and announced.

The registration database is part of an all-encompassing system involving numerous uni- and

bidirectional secure federal, state, and local data feeds as well as from citizens and other

electronic sources. The 2002 Help America Vote Act required states to establish a statewide

VRD and verify voter accuracy by comparing it to other state records.18

The VRD helps estimate voter demand at a precinct allowing officials to adjust staffing levels.

Voters can also check their data for accuracy. As an anti-fraud measure, the database helps

prevent voters from voting twice. Many VRD design goals were specified by the Association for

Computing Machinery to help unify each state’s approach to their system.19

In all jurisdictions, voter registration is an IT function critical to safeguarding free and fair

elections. To ensure democracy’s spirit of “one person, one vote”, each state is authorized to

track voter eligibility. This simple

example shows that New Jersey’s

rules are different from California’s.

Each system requires secure data

feeds from other state or federal

agencies just to maintain these

simple rules.

A great deal of VRD activity is attributed to maintaining eligible voter lists by purging ineligible

citizens through intra-database data exchanges. The process requires automation as some

systems contain millions of voter records. Ineligible names are removed or updated by an

address change in the Post Office National

Change of Address (NCOA) database, state

Department of Motor Vehicle (DMV), or state

tax collection authority. Death notifications,

such as from Florida’s Department of Health and Vital Statistics, or state stipulated murder or

sexual offense disqualifications such as from a Kentucky Department of Corrections data feed

must be tracked. Thirty states also use a data feed from the Electronic Registration Information

Center, a non-profit organization that helps states improve the accuracy of voter rolls.20

As shown by the chart above, flexibility is a critical element of a registration database. The

Federal government can only make recommendations for interoperability, and state and local

administrators must work with other states and municipalities to exchange and incorporate data

New Jersey California

Are a U.S citizen Are a U.S citizen

Are a resident of New Jersey Are a resident of California

Are at least 18 years old by Election Day Are at least 18 years old by Election Day

Live in the precinct where you vote for at

least 30 days prior to the election

You are on parole for a felony conviction

or convicted of a felony

You are on parole for a felony conviction

or convicted of a felony

You have been legally declared "mentally

incompetent" by a court

You are in prison or detention or jail or

penal institution

Example of States’ Rights in Determining Who Can Be In A Voter Database

You are

eligible to

vote in this

state if you:

You are

NOT

eligible to

vote in this

state if you:


from other systems. The nature of the system's uniqueness may dictate that data exchange is

through a physical CD-ROM or DVD, or a communications link. Data exported from a system

must have security controls and imported data may require an Extract, Transform, and Load

(ETL) process if the data is “dirty” or reformatted for database compatibility reasons.

While a common state goal is for seamless data exchange, systems

must allow for non-uniform format ETL data conversion operations

such as when data is exchanged between Florida and New York.

This abbreviated portion of Florida’s file layout to the right shows a

10-character Birth Date MM/DD/YYYY. New York State maintains

the birth date as 8 characters YYYYMMDD.21

Other differences that must be accounted for include Florida’s 30-

character last name while New York uses 50 characters. Common

fields can also have different meanings, allowing a state to define a

unique layout. States can permit voters to declare some of their

information is private

such as their name and

address, while fields

such as gender, race, and party affiliation are deemed

public information. To the left, these two states also

have unique party affiliation abbreviations.

Given the importance of exchanged data to voting integrity, the Federal Government

established a standard NIST data dictionary to assist election IT developers in reducing the data

conversion burden.22 Using a baseline common data format, future enhancements include the

ability to match state driver license numbers. A Federal Unified Markup

Language model defined the necessary Extensible Markup Language (XML) and

JSON schemas to facilitate easier data exchanges such as the AssertionValue

Enumeration and definitions of “no”, “yes”, “unknown”, and “other”.

Adding new voters or updates to database records are triggered by a new registration form,

another database feed such as a released convict that is permitted to vote, a driver passing

their road test and checking off a DMV box requesting registration, driver’s license renewal, and

more. Care is taken to ensure unique entries are maintained and not duplicated. In the data

processing world, algorithms aid list comparison, but the task is still complex. For example,

CPF Constitution Party of Florida BLK No party affiliation

DEM Florida Democratic Party CON Conservative

ECO Ecology Party of Florida DEM Democratic

GRE Green Party of Florida GRE Green

IND Independent Party of Florida IND Independence

LPF Libertarian Party of Florida LBT Libertarian

NPA No Party Affiliation OTH Other

PSL Party for Socialism and Liberation REP Republican

REF Reform Party of Florida SAM Serve America Mvmnt

REP Republican Party of Florida WORWorking Families

Florida New York State

«enumeration»AssertionValue

enumeration literals

noyesunknownother

Field Name Length

Protection

Request

County Code 3

Voter ID 10

Name Last 30 Y

Name Suffix 5 Y

Name First 30 Y

Name Middle 30 Y

Residence Address Line 1 50 Y

Residence City (USPS) 40 Y

Residence State 2 Y

Residence Zipcode 10 Y

Mailing Address Line 1 40 Y

Mailing City 40 Y

Mailing State 2 Y

Mailing Zipcode 12 Y

Gender 1

Race 1

Birth Date 10 Y

Registration Date 10

Party Affiliation 3

Precinct 6 Y

Precinct Group 3 Y

Daytime Phone Number 7 Y

Email address 100 Y

Florida Voter Registration Extract File - Partial


determining that Elizabeth Smith and Betsy Smith (Betsy is a nickname) at these addresses is

or isn’t the same person can require additional

matching such as from the DMV or Social

Security Administration records.

Examples of algorithm matching include a full character name match and Soundex, which is a

phonetic match for names that are pronounced the same. These approaches are not perfect, as

in the case of “Smith”, “Smyth” and “Smythe”, which all share the same Soundex “S530” code.

If a data match cannot be established, human intervention may be needed. In our example,

Betsy at Maple Street may have moved to South Avenue and mistakenly filled out a new voter

application with her nickname instead of a change of address notification with her birth name.

Algorithms can also attempt to find matches based on incomplete information. For instance, the

last name mismatch of Elizabeth Smith at 123 Maple Street could show an NCOA entry for

Jim Kirk at that address, requiring further investigation. Perhaps Elizabeth and Jim both reside

at that address. A match of Elizabeth Smith and Elizabeth J. Smith at that address could

imply one data feed had a middle initial or is a close relative of Elizabeth’s who uses a middle

initial at that address. An administrator can always try to reach out to the voter for clarification.

The same database could be scanned for jury duty candidates, or interstate matching to find

voters registered in two states. It can also assist homeless voters with an entry indicating their

mailing address, such as a relative’s house, is physically different from where they live.

Database matching is a vital administration tool and must precisely follow a rule base. For

example, voter names can have a wide variance, such as a nickname or maiden name on

various forms of identification. Stored signatures must be processed for allowable variances

through alternate signatures, multiple versions of a signature, and illegible signatures since

signature matching is an inexact science. An administrator can often assist a database match

when confusion arises, especially when the administrator brings years of expertise to bear.

Purging Data Records

Record purging can be error-prone given the databases’ volume of daily change. Incorrectly

culling ineligible voters disenfranchises them, while not identifying them threatens voting

integrity, such as if they move to another state and vote in both state elections. Applying

increased eligibility criteria could result in less purged records, while fewer data checks could

purge too many voters. False-positive purges are bad while a false negative purge could

Existing Voter Database Entry New Voter Application

Elizabeth Smith Betsy Smith

123 Maple Street 678 South Avenue

Born: 6/18/64 Born: June 18, 1964

Drivers Lic: 303-2886-97-061864 Social Security: xxx-xx-2239


erroneously keep a citizen on the VRD. To improve transparency, reduce errors, and prevent

unauthorized database access, some states try to notify voters they are being purged.23

Failing to purge records of the dead are part of the alleged 1960 fraud under Chicago Mayor

Daley. Many believe but never proved he arranged for dead voters not yet removed from the

manual system to cast ballots for John Kennedy.24 There were also reports of ballot-box stuffing

under his watch, that in total, allowed Kennedy to win Illinois by a slim 8,858 votes (0.2%).25,26

Automated record purging makes it prudent to have an audit trail and secondary confirmations

through alternate methods. For example, to determine a voter has moved, an administrator

should check with at least two databases such as NCOA and their current state’s DMV. With the

possibility of human error or unauthorized access, VRD data must also be verifiable as well as

encrypted to safeguard privacy for driver's license numbers, birth dates, etc. Voting

commissions should provide tools such as New Jersey’s Division of Elections portal to allow

voters to verify their registration information.27 Independent verification and audit trails aid in

reversing intentional and unintentional entries. Auditable data should be generated any time a

record is created, deleted, or modified, the database undergoes configuration changes, security

policy changes, or the design layout is altered. It helps to have the audit trail in a separate data

medium such as paper or an air-gapped tape backup device (an electronically disconnected and

isolated data copy rather than, for instance, an online cloud backup).

The VRD is a Single Version Of Truth in business management terms – a central database of

every legal voter in a consistent and concise form. Each record follows a published layout and

contains other data that makes it easy for a state to publish public extracts.28

Data from VRDs and other public databases (that may restrict the

use of private information) is turned into an insightful demographic

analysis of how America truly votes. Harvard’s Dataverse extract of

the 2016 U.S. Presidential race shows almost 6% of U.S. voters did

not vote for either Clinton or Trump.29 Over 171,507 voters,

representing 0.125% of total votes, submitted a blank ballot, which

is different than 28,863 voters who selected “None Of The Above”,

152,234 “Other”, 959 for “Over Vote” (when a voter makes more

than one entry per row), 152,493 for “Scattering” (write-in votes for

unregistered candidates), and 963,123 voters than may have chosen a “write-in” candidate.

Voters also picked “James Hedges” 900 times in Colorado and Mississippi while “Jim Hedges”

Candidate Votes Pct

Blank Vote 171,507 0.13%

Castle, Darrell L. 179,096 0.13%

Clinton, Hillary 65,853,581 48.14%

Hedges, James 900 0.00%

Hedges, Jim 4,709 0.00%

Johnson, Gary 4,244,326 3.10%

McMullin, Evan 498,179 0.36%

None Of The Above 28,863 0.02%

Other 152,234 0.11%

Over Vote 959 0.00%

Scattering 152,493 0.11%

Stein, Jill 1,393,155 1.02%

Trump, Donald J. 62,985,062 46.05%

Void Vote 4,278 0.00%

write-in 963,123 0.70%


of Arkansas received 4,709 votes (there is a James “Jim” Hedges who ran for the Prohibition

Party in 2016, but from a legal ballot perspective, the actual name must be precise.)30

Most of the political parties chosen included the popular Democrat, Republican, Independent,

Libertarian, and Green, while some of the parties represented by candidates included “We The

People”, “Approval Voting Party”, “Legal Marijuana Now”, and “Nutrition Party”. Given the 2016

popular vote victory margin between Clinton and Trump was close, and the electoral vote ran

opposite of the popular vote, we can speculate that a pared-down list of available candidates

could have refocused an additional 2.3 million voters and changed the Presidential outcome.

VRD’s use COTS componentry and are subject to the same malware risks and network attack

threats of any desktop computer. Most equipment uses close source code and is subject to non-

disclosure agreements, so it is unclear what the processing algorithms are doing. As more

equipment is added to a network, its potential exposure profile increases. This is especially true

of precincts relying on wireless internet access making encrypted communications mandatory.

Access control and authentication issues can also arise given the number of poll workers that

use the equipment. The nature of distributed networks requires extra care to ensure proper

backups, rollback-recovery, and auditable capabilities are built into the system.

Poll Books

Poll books are created from VRD

extractions just before an election

to create a digital or paper list. The

list allows poll workers to record a

voter signature during the

Election period. Paper records

such as the example to the right

are from a binder of sheets printed for that particular jurisdiction’s polling station. There are

places for the signature, which the poll worker uses to verify against their previously obtained

digitized signature, and address, birth date, and party affiliation which is important during a

primary election. If the voter already submitted an absentee ballot, it would be noted on this

page, preventing them from voting a second time. Using their barcode for the VRD record

identifier, their current live signature can also be captured and likely used for the next election.

An Electronic Poll Book (EPB) displays voter data for a particular station on a COTS console

from a roster of eligible voters downloaded through a wired communications line, WiFi, or a USB


thumb drive. Poll workers could be assisted by algorithmic comparison of the digitized and live

signatures to analyze stylus speed, pressure, and other handwriting

attributes. KNOWiNK’s Poll Pad is an EPB tablet that can add precinct

functions such as same-day registration.31 Poll Pad uses WiFi/MiFi,

AWS’s GovCloud, and a custom database, and is deployed using the

Cisco Meraki MDM that meets FIPS 140-2 and PCI DSS Level 1 security requirements.32

In some states, an EPB must be on a network to receive updates such as who might have voted

in another precinct. In other states such as Michigan, voter data is local to the device and

should not be on a network.33 In general, all networked devices must be part of a secure

communications initiative since it is a point of attack for a cyberterrorist intent on disrupting an

election, possibly enabling a person to impersonate another voter, or serving as an entry point

for a virus that could impact ballot tallies. As with any hardware/software combination, care

would need to be taken that the equipment was not manufactured or updated with a virus. A

computer virus can also infect the VRD through data modification and cause system-wide

damage such as purging of valid voters or changing the final tally.

Casting a Vote

To date, there are four basic methods used to cast a vote. Two involve paper and two use

machines. As with everything discussed, these methods vary by jurisdiction along with the

layout of the paper forms and types of machines.

Paper Ballots – Large heavy-weight paper sheets designed with empty ovals, ellipses, boxes,

and other shapes record a voter’s choice when filled in with a blue

or black pen, felt-tipped pen, or pencil. Any mark not in an oval or box is not read, so a voter

who makes a mistake is unable to cross out or erase an entry, and shapes like crosses and

checkmarks are discouraged. These ballots can be filled out at a polling place or even at home.

In this illustration of an optically read rounded rectangle,

a detectable mark within the red outline can reliably be

processed by the optical scanner as a positive assertion.

Marks that fall outside this shape can yield unacceptable

or unreliable results such as a reliably ignored mark not

detected when

scanned. Marginal marks may or may not be tallied as

intended.


Similar to a barcode reader, Optical Mark Recognition (OMR) uses a photodiode to capture

reflected LED light intensity from the oval’s white or

empty space (“0” or “no”). A black oval absorbs

light, so none is reflected into the photodiode,

which is translated into “1” or “yes”.

Voting categories are arranged into a familiar grid of rows and columns that align with voting

choices. In this hypothetical example from the 2000 Presidential election,

there are three candidates – Al Gore, George Bush, and

Ralph Nader. Each candidate is under their party

headings that align under columns 3, 4, and 5, and the

office they are running for is in row 2.

This ballot page is scanned by OMR that aligns the

page optically as guided by black rectangle timing

marks that intersect at preprinted rows and columns,

ensuring the page is fed straight. If the page is crooked, the ovals will not line up under

photodiodes, and choices will not be sensed and translated correctly.

Using the standard (row, column) array data structure, the three possible Presidential choices

are located at coordinates Ballot (2,3), Ballot (2,4), and Ballot (2,5). In this example, a ballot

passes under a fixed sensor bar, and Ballot (2,3) is translated into a “1” or “yes” for Al Gore,

while Ballot (2,4) and Ballot (2,5) are translated into “0” or “no”. The voter in this example filled

in the oval at address “020311“ as processed by the photodiodes. A page scan creates a

matrix of values that represent the voter’s wishes, and “no” votes can be discarded.

In this way, a standard scanner can be used for a page of names, categories, and questions

that vary by jurisdiction and year. By aligning row and column timing tracks, a program “asks” a

question and simple optics read a voter’s choice. This concept is used for the entire ballot of

officials and questions. A two-sided ballot uses a two-sided scanner.

Paper ballots like these, which may have

been filled out at home weeks before

election day, can be individually scanned at

the polling place or batched processed by a

high-speed scanner at the central election

location. This is an example of a part of a New Jersey paper ballot for the 2020 election cycle.

Column

ADemocratic

President

Al

GORE

Column

BRepublican

George

BUSH

Column

CGreen

Ralph

NADER

Row

2 3 4 5

1

2

Al Gore equalsRow Column Sheet Side02 03 1 1

LED

Phototransistor

Column

2 3 4 51

Fixedsensor

bar

Pathof

paperballot


Paper Punch Cards – Phased out by 2014 because of their inability to reliably capture voter

choices.34 The premise is similar to a paper ballot. A hole at a particular

row and column is translated as a positive selection when the LED light

shining through it is received by a photodiode. If there was no selection

made at that row and column, then there would be no hole and the light

would not be detected by the photodiode, meaning a “0” or “no”.

While simple in concept, what happened in Florida’s Palm Beach County in 2000 was anything

but. Using a punch card User Interface (UI), voters used a pointed tool to make a hole next to

their choice in a “butterfly” ballot. The card was

read by a high-speed reader at the central

location. In this case, the punch card UI failed to

prevent voter confusion and errors. Some voters

ignored the instructions and black location arrow,

making a hole next to both Presidential and Vice-

Presidential names. Some voted for two

candidates - Buchanan and Gore. Others voted for Gore by making a hole next to Buchanan.

Nothing in this poor ballot design would help the voter catch their mistakes.35

Some voters using the tool failed to fully punch out

holes, creating “hanging chad” fragments that fouled

up the automatic card reader. Many punch cards

had to be reviewed by hand, leading to recounts that

went on for weeks as some interpretations were

difficult to make. Bush eventually beat Gore by 537 votes in Florida, marred by an election with

a bad UI that may have changed history and certainly hurt our faith in the voting system.

Direct Recording Electronic (DRE) – A system incorporating a computer and usually a

touchscreen that guided a voter through the voting process, ensuring they cannot vote outside

administrator rules, such as accidentally voting for more than one

Presidential candidate. In some machines, touching the space next

to the candidate highlights the choice in yellow along with a large

green checkmark. DREs became popular, in part, because of

the drawbacks of paper ballots, such as the need to customize, multilanguage versions, large

font versions, printing expenses, distribution to polling places, and storage after the election.


Touchscreens are a key piece of this solution and use resistive or capacitive technology.36

ATMs and some tablets use pressure resistive pads that allow a finger or stylus to compress a

conductive air-gapped plastic against conducting

glass, making it ideal in harsh conditions such as

wintertime when people wear gloves. Capacitive

screens use the fingertip’s skin as a conductor. In either case, the touchscreen surface has a

grid of embedded electrodes. A completed circuit allows the screen controller to process the

coordinates and pass them to the operating system. According to the Voluntary Voting System

Guidelines (VVSG), a touchscreen must be usable by voters with prosthetic devices and not

require direct bodily contact as part of the circuit, and if it does, a stylus must be provided.37

The DRE touchscreen can be multilingual, have different font sizes, and offer audio prompting.

Voter choices are stored in the machine’s memory. When the ballot is submitted, some DREs

print a Voter-Verifiable Paper Trail (VVPT) receipt of selections, similar to a grocery receipt.

When the polling station closes, a supervisor’s master card allows the system to transmit results

to a central site or save the results to removable media such as a USB stick, as well as

generate an administrator VVPT audit printout in the event a recount is needed.

Some DRE systems serve as a Ballot Marking Device (BMD) and print out a marked paper

ballot of the voter’s choices. Using a multilingual UI, along with optional visual or audio prompts,

they can assist with the voting process. DRE technology can also try to understand a voter’s

ballot intentions, and through an improved UI, remove voter confusion and create a positive

experience. For example, a DRE could have prevented a Florida voter from both selecting Joe

Biden for President and writing his name in the “Write-in Candidate” space. Some machines can

create a Quick Response (QR) code or barcode for use by the central processing facility.

Older DREs had lever switches to record votes on

mechanical counters. At election close, the counters

were hand-copied and reported or generated a paper

tape summarizing the counter totals. These machines

may still be in use, but production has ceased and phased out in favor of other solutions.

Barcodes are a key coding technology that improves election speed, accuracy, and efficiency.

Envelopes, paper ballots, VVPT, poll books, and more use barcodes to facilitate processing and

tabulating. Invented in 1952, black and white bars absorb or reflect light into a photodiode using


OMR to generate a data string.38 This barcoded receipt was printed by

a voting machine summarizing the voter’s choices and creating an

auditing barcode of selections as shown inside the red oval.

Similar to Morse code’s dots and dashes, and popular on grocery items, barcodes use a

“1” and “0” code representing light absorbed (black) or reflected (white) stripes. Based on

the code, the data is translated into a meaningful sequence that could contain voter

choices or help postal machines route a ballot envelope to the proper mail carrier’s route.

There are dozens of coding methods, but one of the more interesting ones is

the QR code two-dimensional matrix such as this one.39,40 Look closely and

you will see three alignment squares highlighted in red that orient the QR

code reader. QR codes can encode thousands of characters of data using

the basic photodiode light reflection concept. A BMD’s filled-out printed ballot

often has a QR code that serves as a secondary summary of the voter’s choices. QR codes

also appear on candidate campaign literature allowing voters to scan them with their

smartphone to get more information, such as a candidate’s position on a particular issue.

Verifying the Signature

The coronavirus caused more voters to use mail-in ballots than ever before, moving signature

verification responsibility from the poll worker to the central processing facility. As a result, even

greater reliance on verification was needed to prevent impersonation of a legitimate voter. It is

unlikely the impersonator knew what the previously-stored signatures looked like.

A signature is a key method in determining the voter’s identity, even when a signature changes

over time. Baseline signatures are stored during the registration process or updated during

triggering events. Signatures can also be “versioned” with multiple vintage signatures kept on

file. Over half the states use envelope signature matching to verify a voter’s identity.41

Pretend you are an election official

comparing an envelope signature against

the VRD. Using a signature matching

challenge issued by the NY Times, can

you match two signatures? Make a note of

your choices - the correct answer is found

in the CONCLUSION section of this paper.


With Vote By Mail (VBM) ballots, trained officials compare an envelope signature against the

signature(s) on file. In their judgment, if the two signatures are sufficiently similar, the ballot is

accepted and counted. If the match is in doubt, the ballot is segregated for further consideration

and not counted until the individual’s identity can be verified. At the precinct, poll workers with

typically less signature comparison training make the same judgment call. Years of training

would be needed to turn workers into handwriting experts, so their training reaches a “middle

ground.” They are taught that if there is any doubt, then proceed as if there is no positive match.

At first glance, initial signature screening

seems to work. However, the American Civil

Liberties Union found between 2012 and

2016, racial and ethnic Florida minorities

were more likely to have their mail ballots

rejected for signature issues or a missing inner “secrecy envelope” and not have them cured (a

process to correct a ballot’s signature) when compared to the general voting population.42

Younger voters were about four times as likely to have their ballot rejected and uncured. In

2016, 315,651 mail-in ballots were rejected for a variety of reasons, and with the 2020

projections, the number of rejected ballots could surpass one million.43,44 Signature matching is

a serious election issue, so more care is needed during this phase of vote processing, but

judgment calls can be hard and time-consuming, and lead to increased voter suppression.

When a majority of voters use VBM because of Covid-19, it becomes apparent that manually

comparing signatures is impractical from a time and resource standpoint. As a result, some

jurisdictions are relying on scanned Automated Signature Verification (ASV) to provide a first-

level authentication check. It is similar to the Post Office envelope address-reading system.45

When a neural network algorithm flags a mismatch, a worker manually inspects the signature.46

The voter is contacted with steps to fix the discrepancy, and if there is enough time before the

final ballot count is completed, mismatches can be corrected. Otherwise, the ballot is rejected.

Machine learning algorithms pre-train and numerically

score thousands of genuine and fake signatures,

comparing the envelope signature with the ones on file, all

without political influence.47 Algorithms can adapt to

signatures that change over time, and at the appropriate

confidence level, declare signatures to match. ASV checks take less than a second making it a

valuable VBM tool. There are many signature comparison algorithms, each with strengths and

Age Accepted VBM Rejected VBM Total Accepted VBM Rejected VBM Total

18-21 67,491 (95.8%) 2,941 (4.2%) 70,432 71,374 (96.0%) 2,984 (4.0%) 74,358

22-25 57,903 (96.5%) 2,094 (3.5%) 59,997 82,667 (96.5%) 2,980 (3.5%) 85,647

26-29 93,736 (97.0%) 2,883 (3.0%) 96,619 89,368 (97.2%) 2,558 (2.8%) 91,926

30-44 312,904 (98.4%) 5,030 (1.6%) 317,934 362,017 (98.3%) 6,405 (1.7%) 368,422

45-64 793,996 (99.3%) 5,897 (0.7%) 799,893 887,348 (99.2%) 6,984 (0.8%) 894,332

65+ 1,015,405 (99.5%) 5,088 (0.5%) 1,020,493 1,220,279 (99.5%) 5,796 (0.5%) 1,226,075

Total 2,341,435 (99.0%) 23,933 (1.0%) 2,365,368 2,713,053 (99.0%) 27,707 (1.0%) 2,740,760

2012 General Election 2016 General Election

Florida Elections - Number and Percent of Accepted/Rejected Vote By Mail Ballots by Age


weaknesses. Our signatures constantly change, with features varying such as cursive versus

print, capture pad speed, proportion, spacing, slanted versus straight, and even spelling.48

Graphology became a popular handwriting analysis tool in the 19th century and involves the size

of letters, angles, slopes, spaces, and more.49 Geometrical and analytical analysis are two

examples of algorithms based on graphology that can help compare signatures. Geometrical

verifiers examine the distinctive elements of the

signature of record against the envelope signature

by building and scoring a comparison of three-node

triangles. Similar triangles have a high correlation

score and are likely the same signature. Analytical

analysis tries to find correlations between signature segments as shown by this color-coded

correlation. California officials found no significant increase in ballot rejection using ASV.50

Vote Tabulation

On Election Day, votes in all categories of every jurisdiction need to be tallied. In some

locations, paper ballots are securely transported to a central site, and in other locations,

removable media or printed summaries arrive at town hall for counting. Jurisdictions assemble

the tallies from precincts and determine how many votes each candidate or question received.

Absentee and VBM ballots (these can be the same

terms in some jurisdictions) have their choices

processed by large automated mail sorting

machines that identify envelope thickness, weight,

and voting precinct to begin VRD tracking.51

Based on the jurisdiction, each ballot outer return envelope is barcode scanned by a camera

that can process high volumes of signatures in a

fraction of the time it would take to do manually. In

44 states, signatures are validated by machines

such as this ES&S Mail Ballot Verifier MBV 1000,

which scans and timestamps 100 envelopes per

minute, isolating the signature. Jurisdictions also

use a manual comparison system to display the envelope signature and the VRD’s set of

signatures on a worker’s screen.52,53


After outer envelope verification, the VRD may be updated with a “ballot processed” timestamp.

The envelope then passes through a high-speed opener like this

OMATION 410 to the right that slices a fraction of an inch from the

envelope’s edge. A worker

retrieves an inner envelope,

like the one to the left, from the outer envelope.54

An inner envelope with a voter’s name, address,

and signature is scanned. Betsy Smith’s VRD

entry is updated for a mail-in ballot. At this stage, it is still unknown who she voted for.

In 36 states, voters have a tracking feedback loop, sometimes

through a smartphone app, allowing them to check if their paper

ballot was received and eventually accepted.55

A machine opens the inner envelope and another worker removes the marked anonymous

jurisdiction-specific secret ballot. The ballots are collected, flattened, and sent through an OMR

reader. These votes are added to the jurisdiction, county, and state tallies.

Precinct tabulation can occur with

ballot scanners such as a Hart

Voting Systems eScan or ES&S

M100 as shown. The voter slides

their finished paper ballot into the

ballot entry slot. Devices like the

M100 run BlackBerry Limited’s QNX, an embedded closed source UNIX-like real-time operating

system. 56 Using OMR, both sides of the ballot are simultaneously scanned, and the voter is

alerted to under- and over-voted selections. At election close, it prints candidate and question

tallies and can transmit encrypted results or store them on a PCMCIA memory card.57 The

scanner retains the ballot in a secure lower compartment as part of its audit trail capabilities.

Advanced DREs like this ES&S DS200 let the voter make choices on a touch-

screen without the need for a paper ballot. A barcoded receipt is printed when

the votes are cast. The system has a battery backup, proprietary flash drive,

audit logs, data encryption, and corruption protection through hash code

tabulation.58 Dominion Voting (former Diebold) and Hart Voting Systems offer

similar equipment.


For central processing sites, larger scan and tabulation equipment such as

this ES&S DS850 can full image scan 300 ballots per minute and sort them

into categories such as counted, write-in choices, and needs manual

review (such as when a ballot is not recognized for that precinct.)59

After ballots are tallied, the final election is certified and closed. Official notification of those

candidates who won and lost their election, as well as any passed or declined referendums or

questions is given to officials and the public. Summary totals may be published to a website and

released in formats such as comma-separated values and XML.

The Importance of Auditing

Officials strive for fair elections, yet acknowledge that mistakes happen and fraud exists.

Perfectly honest people make errors, and machines can be misconfigured, have bugs, or be

infected by a computer virus. Administrators rely on independent auditors to ensure election

integrity. An audit determines if people and technology performed their tasks correctly, thereby

reinforcing confidence that election outcomes are legitimate.

The 2000 Presidential election was a textbook auditing example. Tabulation issues, ballot

design, registration, regulations, and operations justified repetitively conducted audits. When

election “fraud” is declared, especially as more technology is applied,

it makes us doubt the process even more. For example, in 2018,

California’s DMV registered new voters through a hacked app like this

one that sent the records to Croatia.60 A VRD log found 100,000

records added to the system, and a software bug added 77,000

duplicate voter records resulting in two registrations for some voters.61

When results are computer recorded and tabulated, it is wise to employ an audit that uses a

different process like a paper trail. Paper trails should be voter-verifiable without reliance on a

suspect machine. In the U.S., 92% of votes cast have a paper record, and paperless DRE

machines are discouraged.62 The IT world uses the same approach – a storage system backup

should be done by a different program and preferably to a medium such as magnetic tape.

An audit should be built into the manual or automated system and be transparent, allowing

interested parties to observe accuracy, note issues, and attempt problem resolution. While rare,

it may be necessary to repeat an audit, so work products should be kept.


Twenty-two states and Washington, D.C. perform automatic recounts when a small margin of

victory is between certain values.63 There are partial and full recounts, with some limited to a

precinct, and others with a statistical Risk-Limiting Audit (RLA) of random ballots to ascertain if

there is evidence of a correct outcome. Recounts reinforce election security and resilience by

allowing people to inspect ballots. For example, undervoted paper ballots with ovals that were

not detected by the scanner, such as this can be

corrected, or in the case of the 2020 election, a Floyd

County, Georgia election official somehow did not

upload votes from a BSD memory card.64 Audits tend to

be unique across jurisdictions and states, just as the

voting process tends to be unique. This 2018 “United

States” chart shows the lack of audit standardization.65

Voter Fraud

Election fraud is probably as old as elections themselves, and a phrase whose meaning

changes over time. Historians have reported that George Washington, well before becoming our

first president, lost his election to the House of Burgesses at the age of 24 with just 7% of the

vote by failing to get voters drunk before the election.66 The phrase “Swilling the planters with

bumbo", meaning supplying the landowners with rum, was a common and clear method of

manipulating an election’s outcome in 1755.67

In the 1860s, William “Boss” Tweed ran a New York City political organization called Tammany

Hall.68 His group was dedicated to having members elected to the NYC government and then

use their political power to enrich the Tammany Hall leaders. One method the Boss employed to

win an election was to get followers to vote multiple times throughout the borough.

Election fraud is a term that encompasses many aspects of intentional corruption of voting laws

and Constitutional amendments, such as the 15th which gave African American men voting

rights, and the 19th allowing women to vote. Briefly, voter fraud is an illegal behavior such as

impersonating another voter to vote twice, selling a vote, a person who votes without the right to

do so, the use of a fraudulent address, and others. Election fraud is illegal election meddling

such as preventing or tampering with voter registration, buying votes, forging candidate petition

signatures, tampering with voting machines, illegal acts by officials to exclude qualified voters,

altering the tabulation and certification of voting results, and more.


The FBI works closely with all government and private officials to disseminate information,

increase security, and stop threats. One of their biggest challenges is to ensure social media,

with its different views of reality, is not used by adversaries trying to circulate fake information.69

Cybercriminals using the GitHub open source Deepfake algorithm published fake social media

videos during the 2020 election to disinform and sway public opinion.70 Deepfakes (Deep

Learning & fake video) use Artificial Intelligence (AI) autoencoder deep learning algorithms to

manipulate videos, making fake events seem real. By encoding images into values and tuning

various parameters, a fake image appears similar to the

original image. In this example, actor Alec Baldwin’s

impersonation of President Trump is “enhanced” with a

deepfake placing the real President’s face on Baldwin.71

Audio can be manipulated to make the deepfake say things the real person never said.

Social media companies like Facebook have policies against fake news. They are creating AI

tools that can be 90% effective in spotting false postings, and with deepfakes, they can key in

on areas such as eye blink rates.72 SybilEdge is a trainable Facebook algorithm that finds troll

fake accounts that pretend to be friendly and connect with real users (friend requests). Vote

trolling starts a quarrel or posts inflammatory comments to decrease trust and sway a vote.73

Some algorithms detect fake news and prevent clickbait (false content causing a user to click its

link). For example, when Donald Trump became the President in 2016, “Guess what????

Donald Trump is the Next US President!!!!!!!!!” was clickbait for a malicious website.74

It can be debated that a confusing ballot design, voter suppression, and asking obtuse ballot

questions is an attempt to commit fraud. Examples include insufficient precinct equipment

resulting in long lines that dissuade voting, and the Florida butterfly ballot.

Technology can mask fraud and errors. In Pennsylvania, Republican Victor Scomillio earned

54,836 votes in his 2019 race for County Judge while his opponent Abe Kassis had 164.75 All

100 precincts used certified ExpressVoteXL DRE machines. Election officials were confused

when some DREs gave Kassis 0 votes in a system where voters can single-click a party line.

Backup paper ballots were checked and showed Kassis beat Scomillio by 1,005 votes, 26,142

to 25,137. What else was wrong? How does the DRE tally show Kassis with 164 while its paper

audit trails gave him 26,142 votes? In the same county, a woman voted straight Democrat and

all Republican candidates lit up, and a man found his choices would not light up.76 Could other

races with this equipment have incorrect totals and faulty audit trails? Was this a configuration


error or hacked equipment? Did the DRE’s self-test check the touchscreen’s code? ES&S later

apologized and said its employees improperly configured the touchscreens and the ballot.

The Pennsylvania incident raises the question – what does voting machine certification mean?

According to the U.S. Election Assistance Commission, this optional certification states a

system was tested by an approved laboratory and meets VVSG requirements and manufacturer

claims.77 It does not necessarily involve configuration details nor security functionality. Machines

are configured long-after testing is complete and security is about preventing an enemy attack.

Experience also shows that any large coding effort will have some number of bugs per thousand

lines of code, and those bugs could permit or foster unforeseen system behavior.

Some elections have the thinnest of winning margins. Any manipulation, whether intentional,

accidental, or an act of God can sway an election. For example, on November 7, 2000, in

Volusia County, Florida, Al Gore had 83,000 votes to Bush’s 62,000. Just 30 minutes later, the

Diebold equipment reduced Gore’s total by 16,000, with the bulk of the difference going to

James Harris, the Socialist Workers Party candidate.78 The problem was eventually traced to a

600 voter precinct and attributed to either a faulty memory card or a phantom second card.79

The Gore – Bush race was very close.80 Needing 270 electoral votes, the polls closed with Gore

at 250 electoral votes to Bush’s 246. Gore got to 268 by winning New Mexico by just 355

popular votes (286,783 to 286,417). On November 8, Bush was believed to win Florida by 1,784

popular votes, a margin triggering a recount. By November 10, Bush was ahead by just 327

popular votes. Recounts continued for weeks. Florida’s Supreme Court and the U.S. Supreme

Court eventually closed the election, with Gore losing and conceding the Presidency to Bush.

Given margins of victory can be small and impacted by fraud, let’s examine areas of election

fraud from a technology standpoint. In 2017, the Department of Homeland Security (DHS)

informed 21 states, including the key electoral vote states of Florida, Ohio, Pennsylvania,

Virginia, and Wisconsin, that “bad actors” from Russia targeted their election systems the

previous year.81 While “targeting” is not the same as “broke into”, in 2016, hackers using illegally

obtained voter registrations for eight states from an election software company sent phishing

emails to over a hundred election officials to try to break into their systems.82 DHS also reported

Russia was scanning computers and networks for security holes.83

There are many bad actors capable of hacking our systems. What are the vulnerable points of

the voting system they could target? Some of the most susceptible entry points include the

VRD, in-precinct check-in, the voting machines, voter tally, and social media attacks:


Susceptible Area #1 – As we’ve seen, the registration database is key to generating mail-in

ballots and populating precinct DRE voting machines. Election systems are widely reported to

be underfunded and often rely on very old equipment, some of which run the registration

database.84 Just like business computers,

the concept of “If it ain’t broke, don’t fix it”

seems to prevail. In 2002, the Help America

Vote Act injected $2 billion into replacing

older machines, but that was almost two

decades ago, and in 2016, the Brennan

Center For Justice found 43 states still had

equipment that was at least 10 years old.85

In general, when old equipment is linked to a complex network, the entire system can become

less secure. A cybercriminal who accesses a system can populate it with fake information and

identities, delete records of their choosing, and generate votes for their candidates. As we’ve

seen, in close elections, a difference of a half-percent could be enough to sway an election.

The Washington Post reported Russian agents penetrated the Democratic National

Committee’s computer network and accessed their database.86 The 2016 Robert Mueller report

to the U.S. Senate titled “Investigation into Russian Interference in the 2016 Presidential

Election” documented Russian access to “each voter's name, address, partial social security

number, date of birth, and either a driver's license number or state identification number”.87

Susceptible Area #2 – States use paper and electronic poll books produced and loaded by the

VRD to log a voter into the correct polling place, verify their signature, enforce any photo ID

requirements, review political party affiliation in the case of a primary, and other functions. EPBs

are an example of precinct equipment that needs secure communications with the central

location. If a voter cannot check-in, they cannot vote. In 2006, Sequoia Voting Systems EPBs,

which is owned by the foreign company Smartmatic, failed voter check-in due to undersized

network issues, high transaction rates, and system uptime/reliability.88,89

Any networked device, such as an EPB, can be a target for a hacker. Hackers could allow

voters and officials to believe a device is correctly recording votes. Even if they do not steal or

change user data or otherwise disable the device, they can trigger a denial-of-service attack

preventing a precinct from servicing voters. Hackers can also target an equipment

manufacturer’s proprietary software, perhaps with a virus that spreads to other devices.


Susceptible Area #3 – Voting machines either optically scan paper ballots or are DRE devices

that print a paper ballot after prompting the voter to make choices. DREs are manufacturer

programmed using non-auditable closed-source code to allow local administrators to use their

VMS to configure devices for each precinct’s common and unique voting choices. The VMS can

leverage USB thumb drives or memory cards which can also contain hacker-provided computer

viruses designed to manipulate votes and tallies. VVPT voter receipts and system paper audit

trails, produced by many DREs, can help combat the manipulation. They can also help hide a

fraud since the paper is printed by the same machine that might have been hacked. A hacked

machine could print anything it wants to hide an attack because secret ballots cannot be linked

back to a particular voter. Voting equipment manufacturers are secretive about not allowing an

independent security evaluation of their machines or the source code running in them.

When the voter signs into a poll book, they are given the next voting number of the day on a

paper slip or a DRE voting card. The voter may also be given a blank ballot or directed to an

available DRE. The DRE’s paper audit trail may be printed sequentially, that is, the first voter on

Election Day has their choices printed at the beginning of the audit trail, followed in order until

the poll closed. If the audit trail is a sequence of voter’s choices, and you were authorized, it is

not hard to break the secrecy and determine exactly how a particular voter voted that day.

There are dozens of studies looking at hacking DREs, and a Google search reveals pages of

hits.90 Some hacks were documented in the 2006 HBO movie “Hacking Democracy” where

candidate Susan Bernecker brought a video camera during her inspection of the DREs stored in

a warehouse before Election Day, 1996.91 She tested the first machine by pressing the button

next to her name and her opponents’ name, Nick Giambelluca, appears in the vote cast display.

She tested 15 machines with identical bad results. Another documentary, “Stealing America:

Vote by Vote”, is a 2008 examination of election manipulation where not all ballots get counted

with claims like “Poll workers watched a hundred and some people go in specifically to that

booth and vote. At the end of the day, when that tape came out, one person had voted.”92

There are also stories about malfunctioning machines, including one that happened to me. A

few years ago, I was casting a vote for my son who was running for local office. I voted straight

party line and could not get the voting machine to put a green X next to his name. I tried

multiple times and it would not illuminate. I finally got the voting machine to capture my vote

correctly, but apparently, I was not alone. After my son won his election, he told me he got fewer

votes than his running mates – that’s when I told him what happened to me. Machines can

malfunction for a host of reasons including the overuse of hand sanitizer during the pandemic.


Susceptible Area #4 – Vote tallying and reporting is performed by closed-source non-audited

systems running on COTS platforms. Until electronic voting systems are auditable, they may be

impossible to secure. Similar to Susceptible Area #3, a hacker may only need to change a

jurisdiction’s tally by a few percentage points to have their chosen candidate win an election.

Beyond a virus, hackers can delay results through denials of service attacks, install

ransomware, and plant social media seeds of doubt into the election’s accuracy.

Business computers can be hacked as well as election equipment. Argonne National Labs

proved a hacker with just $10 in parts and access to a 2012 Diebold

Accuvote TS DRE could alter its functions to change a voter’s choices.93

(NOTE: Diebold sold its election equipment division to ES&S in 2009, which

sold it to Dominion Voting Systems in 2010.94 A portion of the software used in these machines

was written in Serbia.95 Diebold machines are still in use despite their documented problems.96)

Our hacking insight comes from officials like William Evanina, Director of the U.S. National

Counterintelligence and Security Center. He says that foreign adversaries are trying to break

into voting systems, spread disinformation, and trying to collect derogatory information about

campaigns, candidates, and prominent Americans.97 “We are very confident that the election

infrastructure and posture is very resilient. We are not worried about changing votes at scale.

But we are worried about influence on the American voter, and the ability of the American voter

to understand where they should get real information, especially when they are voting. How to

vote. Where to vote. Be patient when you vote. Be prepared.” He said Russia, China, and Iran

are using disinformation campaigns, spreading conspiracies and false information to promote

candidates they favor and sway the 2020 election. Director of National Intelligence John

Ratcliffe said Iran sent threatening

allegedly far-right militia group spoofed

emails, such as this one, to Florida

Republicans using stolen voter data.98

Voters want to trust election systems, yet in Florida, that trust was broken again. Florida officials

had maintained the 2016 election was free from outside attacks. FBI cybersecurity specialists

and the DHS, using the Mueller report, secretly informed 67 county election officials in 2019 that

Russian hackers used "spear-phishing" attacks targeted at election workers with an email link

that downloaded a computer virus.99 The Russians succeeded at hacking into registration data

in at least four Florida counties in 2016, undermining voter trust in Federal officials who kept the

information secret for 3 years. Local election officials were blamed for lax security measures.


Susceptible Area #5 – Unlike filtered and verified information from reliable newspapers and

journalists, some social media sites can knowingly or unknowingly host fake news and AI-

inspired open-source programs like Faceswap, allowing almost anyone to post bogus videos

designed to spread election fiction, false narratives,

and disinform millions of voters.100 Many sites

employ algorithms that keep the reader engaged,

and the Reuters Institute shows 1/3 of Americans

were exposed to fake news during a week in

January/February 2018, stressing election process credibility. The FBI warns that cybercriminals

may be trying to create false social media content and websites, which were not specifically

engineered to ensure truthful information, to spread disinformation, and undermine elections.101

It has become easier for nefarious groups to influence election outcomes by planting doubt and

confusing voters through false truths. Companies like Facebook face an uphill battle to fact-

check propaganda postings and misinformation. Rather than publish content chronologically,

they algorithmically sequence posts and ads based on what they see as relevant to their

audience. During the 2020 election, they even scored the journalism through a News

Ecosystem Quality algorithm and adjusted the content.102 More “likes” and clicks mean more ad

sales. In 2016, Facebook estimates Russian sponsored ads reached 126 million subscribers.103

Future Voting Technology

The history of voting in this country is hundreds of years old, and for the most part, devoid of

modern technology. Before the Revolutionary War of 1775-1783, white

male landowners not caring about anonymity, and sometimes subject

to bribery and intimidation, would yell their votes in public at a carnival

or gathering.104 By the 1800s, voters wrote their names under their

candidate’s name or used pre-voted political party ballots as shown

here that were stuffed into the ballot box.105

Voter privacy became important by 1892 and Grover Cleveland became the first President to be

elected by a secret modern paper “Australian” ballot.106 Sadly, the nature of a secret ballot gave

rise to individuals fraudulently voting more than once.

Lever machines in the 1920s tallied voter’s choices with internal mechanical counters,

advancing the voting process. In 1962, election technology leaped forward when mark-sense

optical scan ballots were introduced. By 1965, voters punched holes in a card next to the


candidate’s name, with the cards centrally tabulated. Punched cards were popular until the

hanging chad debacle of 2000.107 The first computerized video voting terminal, controlled by a

central computer, appeared in 1974.108 In 2002, the Federal Election Commission issued Voting

System Standards about computer-based election system integrity, the same year Georgia

became the first state to use a DRE.109 The first election Hackathon was held in Las Vegas in

2017 and proved that talented engineers could hack into DREs and VRDs in under two hours.110

The goal of future voting technology is to make it easy, secure, auditable, and at a lower cost.

Beginning with online voter registration, citizens should be able to register with a smartphone

app or in trusted institutions that are open evenings and weekends such as a public library, town

hall, or hospital. Citizens could register when starting a new job, through lunch-time office

gatherings, or even allow for automatic registration. Using the same security standards as an

online loan application, an app could guide us through complexities and instantly check our

answers to prevent duplication or conflict when we enter our information. Signatures and photos

could be added if available from the DMV or other state agencies, or updated from a military,

employee, school ID, or even a selfie. This would result in a more complete and accurate VRD

at a lower cost than processing hand-written paper forms.

Voters want to know their choices were secret, which eliminates voter coercion and bribery, and

their intentions were properly tallied. We live in a 24/7 connected world, yet general elections

tend to be limited to the first Tuesday after a November’s Monday, and election authorities view

online voting as inherently dangerous to holding a free and fair election.

To some, losing the ability to observe a citizen voting raises the specter of vote manipulation

and interference from others. They yearn for secure, encrypted, verifiable equipment that is

hardened to attack and can reproduce vote counts based on voter intentions. Paper ballots,

even those generated by BMDs, reflect a century’s old approach to voting and discriminate

against voters with disabilities. Paper audit trails are not ideal but currently a good safeguard

against fraud. They also look towards systems that are simple to maintain, configure, allow for

easier yet reliable registration and data feeds, and help with the accurate VRD purging process.

Voters that deem certain elections as critical are willing to wait in line for early

access or Election Day voting. In Maricopa County, Arizona, the average wait in

2020 exceeded two hours. Wait times are basic queuing theory and directly

related to voter arrival rates, voter resources, and voting duration.111 Waiting for

Wait Time Precincts

0-1 Hour 19

1-2 Hours 9

2-3 Hours 12

3-4 Hours 10

4-5 Hours 5

5+ Hours 5

Precinct Wait Times


hours can be a challenge, especially during inclement weather. Ways to fix it include:

1. Promote early 7-14-day voting that incorporates weekends. 2. Hold elections on November 11th, Veterans Day. It could be a national work-free holiday. 3. Align work hours so voters can access their precinct during election hours. 4. Allow mail-in voting. 5. Allocate more poll workers and machines to geographic areas based on queuing theory. 6. Add photos to the VRD through DMV registration to speed poll book ID verification. 7. Enable DMV “Real ID” gold star smart chip verification as used by the TSA for airlines.

At the polling place, biometric check-in integrity can be achieved through the same facial

recognition system used by the airlines and U.S. customs.112 In place of a verified photo,

fingerprints could be crosschecked against a signature. With state agreement, a Voter ID card

equipped with a barcode or embedded chip could be used during registration and voting.

[NOTE: The topic of a Voter ID card is hotly debated with claims that seven of the thirty-four

states with ID requirements disenfranchise legitimate voters by placing an undue burden on

minorities and other groups given the direct and indirect cost to obtain the ID.113 About 8% of the

population, including 25% of eligible African Americans, do not have a government photo ID.114]

The servers hosting the VDB, processing ballot transactions, and tallying results must use fully

verifiable hardware, software, and communications. Multilingual smart apps could use auditable,

lower-cost open-source programs that software engineers could examine. An open-source

solution would need to establish voter identification, present relevant voter choices, allow the

voter to verify their intent, and produce a bank-level ATM-style transaction along with the

necessary audit trail. Functionality for disabled, hearing, and visually impaired would be

available to these voters through methods they are already familiar with such as Alexa,

headphones, or use white letters on a black background in large fonts. The audit ability is a

critical element of ensuring a fair election, and today, 92% of votes cast have a paper receipt.115

Special hardware, such as a unique touchscreen, should use open-source software and drivers,

with checks that binaries are unaltered and virus-free. While this is counter to the current crop of

proprietary hardware and software solutions, there is no other easy way to ensure computerized

equipment is fraud-free. Open source doesn’t guarantee extra scrutiny, but it provides

transparency and dissuades malicious code. Reproducible security assessments, test scripts,

and scenarios, also part of the public domain, can be used to automatically verify each piece of

technology and communication path before, during, and after an election.

It seems inevitable that technology will continue to advance the state-of-the-art approaches to

voting. Systems will interface with the mobile culture. It defies logic that you can perform a

banking transaction with your Alexa virtual assistant or smartphone but can’t cast a vote.


While the goal of election technology is to make voting easier and more secure, modern history

shows it can be difficult. The 2020 Iowa Democratic caucus met to select Presidential delegates

to the party convention. They used the Shadow company’s IowaReporter smartphone app to let

1,700 precincts take a picture of local results and send them to a central office.116 Unfortunately,

a coding error meant the app couldn’t handle the volume, leaving many users unable to log in,

and those that could found difficulties with the reporting process. The problem caused major

delays in caucus results and left citizens feeling the internet was not ready for the elections. It

showed that proper testing is needed to gain the public’s trust.

While the 2020 election reports showed a greater than 50% voter turnout among young people

18-29, the pattern from recent Census Bureau data shows from 1980-2016,

only twice have more than half of young people voted, far less than other

age groups.117,118 It is largely attributable to voting not being an annual habit,

apathy, protests, registration time, long lines, transportation, work conflicts,

and more. They also found that turnout among those less educated, the

poor, and minorities was also lower than the general population. Some

expressed their apathy by not voting, or not voting for a Presidential choice.

How often have you looked at a live ballot and wondered what a candidate stood for, or if there

was a “Consumer Reports”-style scorecard of how well an incumbent kept their past promises?

Standing in a voting booth under time constraints is not the time to do candidate research. Many

vote a party line because they don’t know individual candidates. If they could vote from home on

their smartphone or even their smart tv, they could take more time to click for trustworthy and

easily understood candidate information before voting. Technology has made everything from

banking to shopping easier in recent years – why not voting? Today, technology can provide the

“plumbing” to address this need, such as BallotReady, which offers information from nearly

twenty categories on any local ballot candidate, such as their views on civil rights, the economy,

and healthcare. BallotReady can even help build a ballot of choices.119

We would all like to be able to find candidates and propositions that truly represent our interests,

however, marrying technology with human nature is difficult. It boils down to politics and biases,

leaving little agreement on impartial factual candidate analysis. While the Democrats point to the

nonpartisan League of Women Voters (LWV) as a trusted source, the Republicans often say the

LWV holds liberal views.120 Until groups reach a consensus or are aided by AI, technology is

limited to present “your side’s” view of a candidate's position. And of course, history and

candidate background is not a promise of how they will legislate if elected.

18-29 30-44 45-64 65+

1980 48.2 67.2 69.8 74.4

1984 49.1 67.1 72.2 75.3

1988 43.8 63.1 72.3 72.7

1992 52.0 67.9 75.1 76.1

1996 39.6 56.9 68.2 69.1

2000 40.3 58.5 67.8 69.6

2004 49.0 62.4 70.4 71.0

2008 51.1 61.8 69.2 70.3

2012 45.0 59.5 67.9 72.0

2016 46.1 58.7 66.6 70.9

Voting Rates by Age: 1980-2016


New Voting Equipment

Given the history of voting, it is encouraging to see new companies introduce ways to solve

election industry issues. Democracy Live is a company whose app, OmniBallot, can use almost

any smartphone, tablet, or PC, allowing a voter to fill out a ballot from wherever they are.121

Designed for the military’s need for absentee ballots, it also helps voters with disabilities and

those living abroad. Used in over 1,000 jurisdictions, including West Virginia, the process begins

by submitting an absentee ballot request form. County officials return a PIN linked to an online

ballot. The voter enters personal information and selects their candidates, enters “write-in”

choices, and picks “yes/no” propositions. Alerts are issued for overvotes and undervotes. The

last step is a printed signature. When done, the ballot is submitted online, emailed, or printed,

and tracked via a website.122 This system is hosted on the FedRAMP-certified AWS cloud.123

In Utah and other jurisdictions, absentee voters use the Voatz app.124 Like

Democracy Live, a voter requests an absentee ballot and installs Voatz

using biometrics and an ID selfie or driver’s license photo scan to display a

ballot. After they make their choices, they securely and privately submit the

ballot using biometric facial recognition credentials. With end-to-end

encryption, Voatz stores information on “multiple, restricted-access,

geographically-distributed servers running on blockchain technology”.125

Paper trails and receipts assure the voter their ballot has been processed.

Smartphone voting promises to streamline the process, encourage higher

voter turnout, and help prevent issues like rainy day hour-long voting lines.

The Intersection of Cryptography and Mathematics May Hold the Answer

A new approach is being developed by Microsoft’s Senior Cryptographer Josh Benaloh to

provide total transparency while giving voters confidence their secret choices are secure and

counted.126 His approach uses end-to-end Homomorphic Encryption (HE). In contrast with the

128-bit Advanced Encryption Standard (AES) which requires decrypting an entire string to be

usable, HE allows encrypted data to be used or manipulated without ever decrypting it. HE

behaves like other public-key encryption methods, but it lends itself to election privacy, patient

medical data processing, search engine privacy, and other technology areas.

With a paper ballot or DRE, you verify your vote and submit it for processing. Some may wonder

if every vote was counted or might have been altered. Ideally, you also want a receipt for your

votes. In this example, Jeff uses HE to guarantee his intentions are counted. Using a BMD,


Jeff’s receipt has his unique full voting string 30487952307 representing his selection of the 3rd

Presidential choice {0,0,1,0}, the 2nd Senatorial

choice {0,1}, and the 1st choice for Sheriff {1,0,0}.

The receipt includes a URL and QR code to his vote

that is stored on a public website that also includes

every vote cast by all voters, including those of Ron,

Brad, Dave, and Sue.127 To start the tally, the 5 sets of votes are multiplied together to form an

encrypted tally 73291066234. When 73291066234 is decrypted, we see the 2nd and 3rd

Presidential entry each had 2 votes {0,2,2,1}, the 1st Senatorial candidate gets 3 votes {3,2},

and the 3rd Sheriff choice wins with 2 votes {1,1,2}. There was no disclosure or alteration of

Jeff’s votes, thereby maintaining secrecy. Auditing these results is public and transparent.

With the URL or QR code, Jeff can see his ciphertext votes are intact and counted as intended.

Unlike AES, no one learns who Jeff or anyone voted for, yet it’s all transparent. It is an end-to-

end verifiable system allowing groups such as the LWV or the Republican National Committee

to quickly reprocess and recount everyone’s vote without divulging anyone’s name. The system

makes it difficult for bad actors to change votes since it is obvious when a vote is altered.

Microsoft incorporated HE into a free, open-source

GitHub software development kit called

ElectionGuard for use with existing voting

equipment.128 As described, ElectionGuard prints a

verification code receipt and a paper ballot for

scanning if a BMD was used or directly with a

DRE. With finalized election results, all HE codes

are publicly published for transparency. To the

right is a representation of Jeff’s ballot flowing

through ElectionGuard once he finalizes his choices and clicks “Print Ballot”.129 To the left, the

ballot scanner reads the ballot to cast the vote in

the precinct or at the central site. Any party can

check if all votes are correctly tallied.

ElectionGuard was piloted in Wisconsin in

February 2020 and is still in development.130


Conclusion

Voting sounds simple. Pose a question, get an answer, and tally the results. Yet history shows it

is anything but simple. America’s election grid is fragmented, underfunded, complicated,

partially staffed by volunteers, occasionally experiences lapses in judgment, burdened by

legislative hurdles, partisan, and troubled by annual fraud accusations.

Despite it all, the election system is a work in progress that functions well enough and generally

delivers free and fair elections. We have reviewed some aspects of voting irregularities and

errors, whether accidental or intentional fraud, and determined that fraud exists and will likely

continue. However, fraud is not rampant and generally doesn’t radically alter the outcome of

elections. The initial review of voter irregularities of the 2020 election cycle concluded it “…was

the most secure in American history.”131 Given the state of the art, the only way to have full

transparency into vote counts and election certification is the burdensome audit of paper ballots.

Nonetheless, America stands for democracy and the idea of free and fair elections. If just one

sentence in the Constitution was amended, election responsibility would fall to the Federal

government instead of each state, shrinking the number of permutations and making it easier for

technology to reshape.132 Every jurisdiction interprets “fairness” differently, leaving voting at a

crossroads. It was common to wait in line to vote, and with Covid-19, we returned to a society

that favors paper ballots and people to deduce if our signature is genuine. Claims of fraud are

still with us, with some attacking the Post Office, others accusing foreign agents of interference,

and some making suggestions on how to vote twice. It is unlikely we have spotted every case of

fraud, but from 2000-2014, there were only 31 impersonation cases out of 1 billion votes cast.133

Do we continue the paper ballot trend? Society relies on the internet and smartphone

transactions, cloud-based AI home speakers to perform banking transactions, GPS directions,

and other disruptive innovations. It seems inevitable that technology, such as homomorphic

encryption, will help reshape America’s election system. Election modernization should allow us

to vote 24/7 weeks before Election Day, provide unbiased insight into choices, and encourage

all eligible Americans to vote.134

Technology can heal political divisions, unify citizens,

and help us select qualified leaders by insisting on

credibility, inclusion, transparency, and accuracy in our

democratic system. As we become smarter voters, we

should hold officials to their oath.


Forward-thinking democracies inevitably need advanced voting infrastructures. But how

precisely and at which degree should they be deployed is complicated. When we think of voting

technology, probably what comes to mind is an app or website you could log onto with a very

clean interface with all the information you’d need and the ability to vote through that interface.

Earlier in the paper, I asked you to pretend you

were an election worker trying to compare envelope

signatures against the VRD using a signature

matching challenge posted by the NY Times. The

answer is number 3 and 9 are the same signature.

If you didn’t pick the matching signatures, then you

would put the corresponding ballot envelope in the mismatched pile and it would require curing.


List of Abbreviations

AES Advanced Encryption Standard

AI Artificial Intelligence

ASV Automated Signature Verification

BMD Ballot Marking Device

COTS Commercial-Off-The-Shelf

DHS Department of Homeland Security

DMV Department of Motor Vehicle

DRE Direct-Recording Electronic

EPB Electronic Poll Book

ETL Extract, Transform, and Load

HE Homomorphic Encryption

LWV League of Women Voters

NCOA Post Office National Change of Address

OMR Optical Mark Recognition

QR Quick Response Code

UI User Interface

VBM Vote By Mail

VMS Voter Management System

VRD Voter Registration Database

VVPT Voter-Verifiable Paper Trail

VVSG Voluntary Voting System Guidelines

XML Extensible Markup Language


Footnotes

1 https://worldpopulationreview.com/state-rankings/number-of-registered-voters-by-state 2 https://worldpopulationreview.com/us-city-rankings/how-many-cities-are-in-the-us 3 https://sharetngov.tnsosfiles.com/sos/election/minutes/20151201_SECMinutes_VotingMachine.pdf 4 https://theamericanleader.org/storylines/expansion-of-voting-rights/ 5 https://youtu.be/pI_t66oV6-M 6 https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-8497.1967.tb00802.x 7 https://learn.cisecurity.org/CIS-Elections-eBook-15-Feb-pdf 8 “Hart Voting System Support Procedures Training Manual”, https://sos.idaho.gov/elect/Clerk/Hart/ac6300-006_62D_SupportProcedures_%23390-cp.pdf, P. 7 9 https://www.sos.state.tx.us/elections/forms/sysexam/voting-sys-bycounty.pdf 10 “Administering Elections: How American Elections Work”, ISBN 978-1-349-55293-1, P. 3 11 https://constitution.congress.gov/constitution/ 12 https://www.pewresearch.org/fact-tank/2020/11/03/in-past-elections-u-s-trailed-most-developed-countries-in-voter-turnout/ 13 https://www.eac.gov/statewide-voter-registration-systems 14 https://www.ncsl.org/research/elections-and-campaigns/election-administration-at-state-and-local-levels.aspx 15 https://ballotpedia.org/Voting_methods_and_equipment_by_state 16 “Asking the Right Questions About Electronic Voting” by Richard Celeste, Dick Thornburgh, and Herbert Lin ISBN 978-0-309-10024-3 17 https://news.yahoo.com/mail-votes-could-delay-election-120337599.html 18 https://www.govinfo.gov/content/pkg/PLAW-107publ252/html/PLAW-107publ252.htm 19 https://www.acm.org/binaries/content/assets/public-policy/usacm/e-voting/reports-and-white-papers/vrd_report2.pdf 20 https://ericstates.org/ 21 https://www.elections.ny.gov/NYSBOE/Forms/FOIL_VOTER_LIST_LAYOUT.pdf 22 https://pages.nist.gov/VoterRecordsInterchange/ 23 https://www.acm.org/binaries/content/assets/public-policy/usacm/e-voting/reports-and-white-papers/vrd_report2.pdf 24 https://nypost.com/2000/11/10/ironic-twist-daleys-dad-helped-steal-vote-for-jfk/ 25 https://www.dnainfo.com/chicago/20161019/downtown/vote-rigged-elections-history-fraud-stolen-trump/ 26 https://www.washingtonpost.com/news/monkey-cage/wp/2017/08/08/heres-a-voter-fraud-myth-richard-daley-stole-illinois-for-john-kennedy-in-the-1960-election/ 27 https://voter.svrs.nj.gov/registration-check 28 https://dos.myflorida.com/media/696057/voter-extract-file-layout.pdf 29 https://dataverse.harvard.edu/file.xhtml?persistentId=doi:10.7910/DVN/42MVDX/MFU99O&version=5.0 30 https://en.wikipedia.org/wiki/James_Hedges 31 https://knowink.com/product-catalog/poll-pad/ 32 https://votingsystems.cdn.sos.ca.gov/vendors/knowink/ki-2-5-0-sec.pdf 33 https://www.ncsl.org/research/elections-and-campaigns/electronic-pollbooks.aspx 34 https://www.pewresearch.org/fact-tank/2016/11/08/on-election-day-most-voters-use-electronic-or-optical-scan-ballots/ 35 “Human-Computer Interaction and The User Interface”, P. 7, 2018 Dell EMC Knowledge Sharing Article 36 https://www.idt.com/us/en/document/atc/newelectronics-capacitivetouchscreens-game-changing-technology-jan2011 37 https://www.eac.gov/sites/default/files/TestingCertification/2020_02_29_vvsg_2_draft_requirements.pdf 38 www.barsnstripes.com/docs/retailbarcodes.pdf 39 http://en.wikipedia.org/wiki/QR_code 40 https://electionconnect.com/cast-your-vote/ 41 https://www.nytimes.com/interactive/2020/10/07/upshot/mail-voting-ballots-signature-matching.html 42 https://www.aclufl.org/sites/default/files/aclufl_-_vote_by_mail_-_report.pdf 43 https://www.newsweek.com/democrats-reliance-easily-disqualified-mail-ballots-could-cost-them-opinion-1538377 44 https://www.usatoday.com/in-depth/news/investigations/2020/10/08/rejected-mail-ballots-projected-major-factor-2020-election/3576714001/

https://www.eac.gov/statewide-voter-registration-systems


45 https://pe.usps.com/businessmail101?ViewName=DeliveryAddress 46 https://www-cdn.law.stanford.edu/wp-content/uploads/2020/04/SLS_Signature_Verification_Report-5-15-20-FINAL.pdf 47 https://www.parascript.com/wp-content/uploads/2020/05/SignatureXpert-for-VBM-brochure.pdf 48 https://www.sos.state.co.us/pubs/elections/docs/SignatureVerificationGuide.pdf 49 https://www.bbc.com/news/magazine-22198554 50 https://www-cdn.law.stanford.edu/wp-content/uploads/2020/04/SLS_Signature_Verification_Report-5-15-20-FINAL.pdf 51 https://www-cdn.law.stanford.edu/wp-content/uploads/2020/04/SLS_Signature_Verification_Report-5-15-20-FINAL.pdf 52 Connecticut, District of Columbia, Iowa, Maryland, New Mexico, Vermont and Wyoming do not perform signature verification. https://www.ncsl.org/research/elections-and-campaigns/absentee-and-early-voting.aspx 53 https://www.essvote.com/blog/video/video-mail-ballot-verifier/ 54 https://www.quadient.com/en-AU/mail/document-handling-equipment/omation-306 55 https://www.realclearpolitics.com/articles/2020/05/06/there_were_not_28_million_missing_mail-in_ballots_143123.html#! 56 https://en.wikipedia.org/wiki/QNX 57 https://www.usenix.org/legacy/event/evt08/tech/full_papers/aviv/aviv.pdf 58 https://www.essvote.com/wp-content/uploads/2020/08/DS200_One-Sheet.pdf 59 https://www.essvote.com/wp-content/uploads/2020/09/DS850_One-Sheet.pdf 60 https://www.latimes.com/politics/la-pol-ca-california-motor-voter-problems-investigation-20190409-story.html 61 https://www.latimes.com/politics/la-pol-ca-motor-voter-registrations-errors-20180524-story.html 62 https://www.fbi.gov/video-repository/interagency-election-security-psa-100520.mp4/view 63 https://www.ncsl.org/research/elections-and-campaigns/automatic-recount-thresholds.aspx 64 https://www.ajc.com/politics/georgia-recount-uncovers-2600-new-votes-in-presidential-race/I75NSPYYGNF43HQZBPYKJWJ5MA/ 65 http://electionlab.mit.edu/sites/default/files/2019-06/Election-Auditing-Key-Issues-Perspectives.pd 66 https://en.wikipedia.org/wiki/Electoral_history_of_George_Washington 67 https://www.smithsonianmag.com/arts-culture/swilling-the-planters-with-bumbo-when-booze-bought-elections-102758236/ 68 https://www.fasttrackteaching.com/ffap/Unit_4_Cities/U4_Tammany_Hall_NYC.html 69 https://www.fbi.gov/video-repository/interagency-election-security-psa-100520.mp4/view 70 https://github.com/iperov/DeepFaceLab 71 https://thenextweb.com/artificial-intelligence/2018/02/21/deepfakes-algorithm-nails-donald-trump-in-most-convincing-fake-yet/ 72 https://www.theverge.com/platform/amp/2019/6/27/18715235/deepfake-detection-ai-algorithms-accuracy-will-they-ever-work 73 https://research.fb.com/blog/2020/04/detecting-fake-accounts-on-social-networks-with-sybiledge/ 74 https://www.sciencedirect.com/science/article/pii/S1877050918318210 75 https://www.nytimes.com/2019/11/30/us/politics/pennsylvania-voting-machines.html 76 https://www.mcall.com/news/elections/mc-nws-northampton-county-elections-complaints-20191227-rlm547dt7raszogqwxyenu2sh4-story.html 77 https://www.eac.gov/voting-equipment/frequently-asked-questions 78 https://www.nytimes.com/2000/11/10/us/2000-campaign-florida-vote-democrats-tell-problems-polls-across-florida.html 79 https://www.nybooks.com/daily/2018/11/05/voting-machines-what-could-possibly-go-wrong/ 80 https://en.wikipedia.org/wiki/2000_United_States_presidential_election 81 https://www.nytimes.com/2017/09/22/us/politics/us-tells-21-states-that-hackers-targeted-their-voting-systems.html 82 https://www.nytimes.com/2017/09/22/us/politics/us-tells-21-states-that-hackers-targeted-their-voting-systems.html 83 https://www.technologyreview.com/2018/08/15/141028/four-big-targets-in-the-cyber-battle-over-the-us-ballot-box/ 84 https://www.pewtrusts.org/en/research-and-analysis/blogs/stateline/2020/07/08/election-experts-warn-of-november-disaster 85 https://www.brennancenter.org/our-work/research-reports/americas-voting-machines-risk


86 https://www.washingtonpost.com/world/national-security/russian-government-hackers-penetrated-dnc-stole-opposition-research-on-trump/2016/06/14/cf006cb4-316e-11e6-8ff7-7b6c1998b7a0_story.html 87 https://cdn.cnn.com/cnn/2019/images/04/18/mueller-report-searchable.pdf, P. 22 88 https://www.washingtonpost.com/wp-dyn/content/article/2006/10/30/AR2006103001224.html 89 https://www.coloradoindependent.com/2006/11/26/poll-worker-sequoia-to-blame-not-user-error/ 90 https://jhalderm.com/pub/papers/diebold-ttbr07.pdf https://security.cs.georgetown.edu/~msherr/papers/sequoia.pdf https://people.eecs.berkeley.edu/~daw/papers/sarasota07.pdf 91 “Hacking Democracy” https://youtu.be/6YldIdkjrqM timestamp 5:36 92 https://www.stealingamericathemovie.org/index.html 93 https://www.computerworld.com/article/2511508/argonne-researchers--hack--diebold-e-voting-system.html 94 https://en.wikipedia.org/wiki/Premier_Election_Solutions 95 https://en.wikipedia.org/wiki/Dominion_Voting_Systems 96 https://www.votetexas.gov/mobile/voting/systems/accuvote.htm 97 “Today Show” October 7, 2020 98 https://www.wuft.org/news/2020/10/20/fbi-investigating-threatening-emails-sent-to-democrats-in-florida/ 99 https://www.sun-sentinel.com/opinion/editorials/fl-op-edit-russia-hack-florida-elections-secrecy-20191028-xrfjiq4vbvelbfwx7chbtlanki-story.html 100 https://www.weforum.org/agenda/2020/10/deepfake-democracy-could-modern-elections-fall-prey-to-fiction/ 101 https://www.ic3.gov/Media/Y2020/PSA200928 102 https://www.nytimes.com/2020/11/24/technology/facebook-election-misinformation.html 103 https://www.npr.org/2019/04/24/716374421/fact-check-russian-interference-went-far-beyond-facebook-ads-kushner-described 104 https://time.com/4305508/paper-ballot-history/ 105 https://www.history.com/topics/us-presidents/vote-tech-video 106 https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-8497.1967.tb00802.x 107 https://votingmachines.procon.org/historical-timeline/ 108 https://patents.google.com/patent/US3793505A/en 109 https://votingmachines.procon.org/wp-content/uploads/sites/46/vss2002.pdf 110 https://fortune.com/2017/07/31/defcon-hackers-us-voting-machines/ 111 https://www.brennancenter.org/sites/default/files/analysis/Long_Voting_Lines_Explained.pdf 112 https://www.cbp.gov/travel/biometrics/biometric-exit-faqs 113 https://www.aclu.org/other/oppose-voter-id-legislation-fact-sheet 114 https://www.coursera.org/lecture/digital-democracy/voter-authentication-q8d1V 115 https://www.fbi.gov/video-repository/interagency-election-security-psa-100520.mp4/view 116 https://www.theatlantic.com/technology/archive/2020/02/iowa-caucus-app-tech/606094/ 117 https://www.census.gov/newsroom/blogs/random-samplings/2017/05/voting_in_america.html 118 https://www.nbcnews.com/politics/2020-election/graphic-battleground-state-turnout-2020-election-n1247337 119 BallotReady.org 120 https://www.insidernj.com/inside-league-women-voters-flap-gop/ 121 https://democracylive.com/omniballot-online/ 122 https://www.youtube.com/watch?v=Thqjdb50TGM&feature=youtu.be 123 https://democracylive.com/wp-content/uploads/2020/04/OmniBallot-Fact-Sheet-Democracy-Live-AWS_3.30.20.pdf 124 https://voatz.com/how-it-works/ 125 https://voatz.com/security-and-technology/ 126 https://builtin.com/cybersecurity/electionguard-homomorphic-encryption 127 https://youtu.be/BYRTvoZ3Rho 128 https://github.com/microsoft/electionguard 129 https://github.com/microsoft/electionguard/blob/main/docs/guide/Verifiable_Election.md 130 https://blogs.microsoft.com/on-the-issues/2020/02/17/wisconsin-electionguard-polls/ 131 https://www.cisa.gov/news/2020/11/12/joint-statement-elections-infrastructure-government-coordinating-council-election 132 https://www.senate.gov/artandhistory/history/common/contested_elections/election_laws.htm 133 https://www.brennancenter.org/our-work/research-reports/10-voter-fraud-lies-debunked 134 https://www.fairvote.org/voter_turnout#voter_turnout_101

https://cdn.cnn.com/cnn/2019/images/04/18/mueller-report-searchable.pdf

https://jhalderm.com/pub/papers/diebold-ttbr07.pdf

https://security.cs.georgetown.edu/~msherr/papers/sequoia.pdf


Dell Technologies believes the information in this publication is accurate as of its publication

date. The information is subject to change without notice.

THE INFORMATION IN THIS PUBLICATION IS PROVIDED “AS IS.” DELL TECHNOLOGIES

MAKES NO RESPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO

THE INFORMATION IN THIS PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED

WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Use, copying and distribution of any Dell Technologies software described in this publication

requires an applicable software license.

Copyright © 2021 Dell Inc. or its subsidiaries. All Rights Reserved. Dell Technologies, Dell, EMC, Dell EMC and other trademarks are trademarks of Dell Inc. or its subsidiaries. Other trademarks may be trademarks of their respective owners.

Documents

CAN TECHNOLOGY RESHAPE AMERICA’S ELECTION SYSTEM?